Yi-1.5-9B-Chat-16K

MERA Created at 20.09.2024 06:34
0.399
The overall result
325
Place in the rating
Weak tasks:
204
RWSD
329
PARus
229
RCB
332
ruEthics
347
MultiQ
344
ruWorldTree
376
ruOpenBookQA
437
CheGeKa
342
ruMMLU
126
ruHateSpeech
338
ruDetox
280
ruHHH
339
ruTiE
292
USE
263
MathLogicQA
272
ruMultiAr
178
SimpleAr
405
LCS
257
BPS
191
ruModAr
379
MaMuRAMu
265
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.086 Accuracy
RCB 0.537 / 0.494 Accuracy F1 macro
USE 0.133 Grade norm
RWSD 0.558 Accuracy
PARus 0.718 Accuracy
ruTiE 0.597 Accuracy
MultiQ 0.308 / 0.192 F1 Exact match
CheGeKa 0.023 / 0.007 F1 Exact match
ruModAr 0.51 Exact match
MaMuRAMu 0.52 Accuracy
ruMultiAr 0.255 Exact match
ruCodeEval 0.002 / 0.009 / 0.018 Pass@k
MathLogicQA 0.415 Accuracy
ruWorldTree 0.764 / 0.764 Accuracy F1 macro
ruOpenBookQA 0.643 / 0.642 Accuracy F1 macro

Evaluation on open tasks:

Go to the ratings by subcategory

The table will scroll to the left

Task name Result Metric
BPS 0.937 Accuracy
ruMMLU 0.475 Accuracy
SimpleAr 0.979 Exact match
ruHumanEval 0 / 0 / 0 Pass@k
ruHHH 0.646
ruHateSpeech 0.796
ruDetox 0.142
ruEthics
Correct God Ethical
Virtue 0.254 0.306 0.186
Law 0.242 0.295 0.166
Moral 0.288 0.324 0.193
Justice 0.232 0.258 0.163
Utilitarianism 0.218 0.285 0.176

Information about the submission

Mera version
v.1.2.0
Torch Version
2.3.1
The version of the codebase
aec92f8
CUDA version
12.1
Precision of the model weights
float16
Seed
1234
Batch
1
Transformers version
4.42.3
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Yi-1.5-9B-Chat-16K

Model size

8.8B

Model type:

Opened

SFT

Architecture description:

The Yi series models adopt the same model architecture as Llama but are NOT derivatives of Llama.

Description of the training:

Yi has independently created its own high-quality training datasets, efficient training pipelines, and robust training infrastructure entirely from the ground up.

Pretrain data:

Yi-1.5 is an upgraded version of Yi (which was trained on 3T multilingual corpus). It is continuously pre-trained on Yi with a high-quality corpus of 500B tokens and fine-tuned on 3M diverse fine-tuning samples.

License:

apache-2.0

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=["\n"]; \nchegeka - do_sample=false;until=["\n"]; \nrudetox - do_sample=false;until=["\n"]; \nrumultiar - do_sample=false;until=["\n"]; \nuse - do_sample=false;until=["\n","."]; \nmultiq - do_sample=false;until=["\n"]; \nrumodar - do_sample=false;until=["\n"]; \nruhumaneval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6; \nrucodeeval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6;

The size of the context:
simplear, bps, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, rutie, ruhumaneval, rucodeeval - 16384

Ratings by subcategory

Metric: Grade Norm
Model, team 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 8_0 8_1 8_2 8_3 8_4
Yi-1.5-9B-Chat-16K
MERA
0.2 0.233 0.6 0.1 0 0.067 0 - 0 0 0 0.067 0.2 0 0.1 0.333 0.033 0.033 0 0 0 0.3 0.1 0.033 0 0.225 0.133 0.033 0.3 0.167 0.267
Model, team Honest Helpful Harmless
Yi-1.5-9B-Chat-16K
MERA
0.541 0.61 0.793
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Management Philosophy Prehistory Human aging Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Biology (college) Physics (college) Human Sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine learning Medical genetics Professional law PR Security studies Chemistry (школьная) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual_physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) European history Government and politics
Yi-1.5-9B-Chat-16K
MERA
0.378 0.398 0.493 0.705 0.507 0.692 0.66 0.514 0.5 0.498 0.447 0.389 0.39 0.593 0.51 0.549 0.64 0.41 0.311 0.458 0.27 0.398 0.37 0.41 0.455 0.44 0.381 0.537 0.706 0.31 0.69 0.653 0.552 0.626 0.464 0.487 0.44 0.532 0.377 0.394 0.606 0.338 0.503 0.533 0.545 0.426 0.583 0.433 0.39 0.42 0.41 0.637 0.523 0.513 0.71 0.655 0.497
Model, team SIM FL STA
Yi-1.5-9B-Chat-16K
MERA
0.581 0.506 0.6
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Managment Philosophy Pre-History Gerontology Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Bilology (college) Physics (college) Human sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine Learning Genetics Professional law PR Security Chemistry (college) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical Engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional Accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) Europe History Government and politics
Yi-1.5-9B-Chat-16K
MERA
0.311 0.465 0.45 0.509 0.5 0.569 0.448 0.333 0.423 0.492 0.615 0.567 0.375 0.535 0.386 0.284 0.654 0.489 0.404 0.649 0.333 0.559 0.6 0.45 0.689 0.455 0.615 0.474 0.807 0.533 0.778 0.564 0.464 0.544 0.439 0.571 0.578 0.578 0.456 0.462 0.473 0.413 0.733 0.711 0.828 0.8 0.466 0.818 0.554 0.719 0.778 0.362 0.709 0.597 0.512 0.363 0.611
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat-16K
MERA
0.254 0.242 0.288 0.232 0.218
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat-16K
MERA
0.306 0.295 0.324 0.258 0.285
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat-16K
MERA
0.186 0.166 0.193 0.163 0.176
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-9B-Chat-16K
MERA
0.778 0.771 0.765 0.757 0.714 0.885